Multi-loop cable group electromagnetic loss intelligent calculation method and device

Abstract

The invention relates to a multi-loop cable group electromagnetic loss intelligent calculation method and device, and belongs to the technical field of power cable operation, and the method comprises the steps: building a cable group electromagnetic field model containing multiple loops through a finite element method, randomly generating a current sample, calculating the loss of a corresponding cable core and a metal sleeve, and forming a current-loss sample set; constructing a GWO-RBF-Adaboost intelligent prediction model, performing global optimization on the central point, the width and the output weight of a radial basis function neural network RBF by using a grey wolf optimization algorithm GWO, taking the RBF neural network as a basis learning device of Adaboost, and training by using a current-loss sample set to obtain a current-loss mapping relation model; and inputting current data of a multi-loop cable group to be evaluated into the trained GWO-RBF-Adaboost model, and predicting cable core loss and metal sleeve loss of each loop. According to the method, the electromagnetic loss of the multi-loop cable group is conveniently calculated through the current, and the calculation precision of the electromagnetic loss is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power cable operation, in particular to a multi-loop cable group electromagnetic loss intelligent calculation method and device. BACKGROUND

[0002] Accurate evaluation of the cable core temperature and dynamic current-carrying capacity of the power cable group is a key supporting technology for modern power system emergency load, transferred load and new energy access. Researchers have proposed various thermal circuit models to support convenient and rapid evaluation of the dynamic current-carrying capacity of the power cable group. The heat sources of the existing thermal circuit model include electromagnetic losses of the cable core and metal sleeve and insulation layer losses.

[0003] The insulation layer loss is related to the cable voltage and is not affected by adjacent cables. There are related calculation formulas. The electromagnetic losses of the cable core and metal sleeve of the power cable group are greatly affected by adjacent cables. The existing power cable core and metal sleeve loss calculation methods include IEC60287 and numerical calculation methods.

[0004] IEC60287 only provides loss calculation formulas for simple single-loop and double-loop power cables. For tunnel and cable bridge and other cable groups containing more loops, only the cable core and metal sleeve loss calculation formulas provided by IEC60287 can be used to calculate the loss of adjacent loops, and the influence of cables far away is ignored. Obviously, the calculation result has deviation.

[0005] The numerical calculation method can calculate the power cable group with arbitrary arrangement and arbitrary number of loops, and has high calculation accuracy. However, considering the thermal-electric coupling, and the existing calculation depends on ANSYS / COMSOL / MAGNET / OPERA and other tool software, it is impossible to directly calculate the power cable core temperature and dynamic current-carrying capacity with the thermal circuit model in engineering.

[0006] Therefore, it is necessary to explore a convenient and rapid calculation method for calculating electromagnetic losses of multi-loop cable groups from currents, so as to be coupled with the thermal circuit model to directly calculate the cable core temperature of the power cable, and then calculate the dynamic current-carrying capacity by using an iterative method. SUMMARY

[0007] In view of the above analysis, the present application aims to disclose a multi-loop cable group electromagnetic loss intelligent calculation method and device, which solves the problem of multi-loop cable group electromagnetic loss calculation.

[0008] In one aspect, the present application discloses a multi-loop cable group electromagnetic loss intelligent calculation method, comprising: Step S1, establishing a cable group electromagnetic field model containing multiple loops by using the finite element method, randomly generating current samples and calculating the corresponding cable core and metal sleeve losses to form a current-loss sample set; Step S2: Construct the GWO-RBF-Adaboost intelligent prediction model. Use the Grey Wolf Optimization Algorithm (GWO) to globally optimize the center point, width, and output weights of the Radial Basis Network (RBF). Use the RBF neural network as the base learner of Adaboost and train the current-loss mapping relationship model using the current-loss sample set. Step S3: Input the current data of the multi-circuit cable group to be evaluated into the trained GWO-RBF-Adaboost model to predict the core loss and sheath loss of each circuit.

[0009] Furthermore, in step S1, in the established electromagnetic field model of the cable group containing ≥3 loops, the Poisson equation is used for the conductive region and the Laplace equation is used for the non-conductive region, and the boundary condition at infinity is applied. Through multiple randomly generated current input samples, the eddy current loss of the corresponding cable core and metal sheath is calculated to form a current-loss sample set.

[0010] Further, step S1 includes: Step S101: Establish vector magnetic potential equations for each region of the cable field, with vector magnetic potential as the independent variable, including the Poisson equation for the conductive region, the Laplace equation for the non-conductive region, and the boundary condition at infinity. Step S102: Introduce the Coulomb specification to obtain the finite element matrix equation, calculate the vector magnetic potential of each node, and derive the eddy current density of the cable core region composed of the conductor, insulation layer and metal shielding layer. Step S103: Calculate the unit electromagnetic loss; the unit electromagnetic loss includes the loss of the cable core region calculated based on the eddy current density and the source current density, and the loss of the metal sheath region calculated based on the eddy current density. Step S104: Randomly generate multiple sets of loop current values, repeat step S103 to calculate the cable core loss and metal sheath loss corresponding to each set of current values ​​as current-loss samples, and form a current-loss sample set.

[0011] Furthermore, the GWO-RBF-Adaboost training process includes: Step S201: Read in the current-loss training set; Step S202: Initialize the gray wolf population; each wolf corresponds to a set of RBF center, width, and output weights; Step S203: Perform iterative optimization; In each iteration, an RBF weak learner is generated using the current wolf parameters and the Adaboost prediction error is calculated. The wolf pack position is updated by α, β, and δ wolves with the reciprocal of the error as the fitness. At the same time, Adaboost adjusts the weights of the samples and the weak learner until the iteration or error termination condition is met. Step S204: Output the GWO-RBF-Adaboost model corresponding to the optimal wolf parameters.

[0012] Furthermore, the iterative optimization process includes: Step S2031: Calculate model error; Use the RBF network parameters corresponding to the current gray wolf position to train the Adaboost ensemble model, and calculate the prediction error on the training set; Step S2032: Calculate fitness; Calculate the fitness value of each individual gray wolf based on the model error. The smaller the error, the higher the fitness. Step S2033: Determine whether the convergence condition is met. If not, proceed to step S2034; if yes, proceed to step S2035. Step S2034: Update the gray wolf's position and return to step S2031 to continue model training; Based on the hierarchical hunting mechanism of the gray wolf optimization algorithm, the positions of ω wolf are updated by α, β, and δ wolves, and the RBF network parameters are adjusted accordingly. Step S2035: Evaluate the model performance at the current location of the gray wolf, and record the individual errors and the overall prediction error after integration; Step S2036 determines whether the number of iterations or the convergence condition has met the requirements; if yes, output the GWO-RBF-Adaboost model corresponding to the optimal wolf parameters; otherwise, proceed to step S2037. Step S2037: Adjust sample weights according to the error; The Adaboost mechanism adjusts the sample weights according to the prediction error, so that the subsequent weak learner pays more attention to the samples that are difficult to fit, and then returns to step S2031.

[0013] Furthermore, the update formula for updating the position of wolf ω, guided by α, β, and δ wolves, is as follows: ; in, ; ; ; This is the current position of Gray Wolf ω. This is the update location for Grey Wolf ω; , , Let α, β, and δ represent the current positions of the wolves. , , , , , Let be the synergy coefficient; where, ; ; ; ; ; ; It is a random vector; It decreases linearly from 2 to 0.

[0014] Furthermore, the Adaboost mechanism adjusts sample weights based on prediction error, the process of which includes: 1) Calculate the normalized weighted error of the current weak learner; ; in For the first The weights of each sample; ; 2) Calculate the ensemble weights of the weak learner; Weak learner weights , The learning rate; 3) Update the sample weights; ; And for all Renormalization, This allows for greater focus on samples that are difficult to fit in the next iteration.

[0015] Furthermore, when constructing the electromagnetic field model of the cable group, the cable core and metal sleeve are set as single coils, and each loop cable core is input with three-phase AC currents that are 120 degrees out of phase. According to the cable grounding method, when the grounding is single-end, the current of the metal sleeve is set to 0, and when the grounding is double-end, the metal sleeve is not set separately.

[0016] Furthermore, the predicted core loss and sheath loss of each circuit in step S3 are used as the heat source for the thermal circuit model to calculate the cable core temperature, and further to evaluate the dynamic current carrying capacity and provide early warning of the operating status.

[0017] This invention also discloses an intelligent calculation device for electromagnetic losses in multi-circuit cable groups, used to implement the intelligent calculation method for electromagnetic losses in multi-circuit cable groups as described above, comprising: The sample set construction module is used to establish an electromagnetic field model of a cable group containing multiple loops using the finite element method, randomly generate current samples and calculate the corresponding cable core and metal sheath losses to form a current-loss sample set. The model training module is used to construct the GWO-RBF-Adaboost intelligent prediction model. The center point, width and output weight of the radial basis function neural network RBF are globally optimized using the Grey Wolf Optimization Algorithm (GWO). The RBF neural network is used as the base learner of Adaboost, and the current-loss mapping relationship model is obtained by training with the current-loss sample set. The loss calculation module is used to input the current data of the multi-circuit cable group to be evaluated into the trained GWO-RBF-Adaboost model to predict the core loss and sheath loss of each circuit.

[0018] This invention can achieve one of the following beneficial effects: This invention discloses an intelligent calculation method and device for electromagnetic losses in multi-circuit cable groups, enabling convenient and rapid calculation of electromagnetic losses in multi-circuit cable groups. The established GWO-RBF-Adaboost model, through the synergistic effect of GWO, RBF, and Adaboost, significantly improves the model's prediction accuracy, generalization ability, and robustness. The solution greatly improves the automation and intelligence level of cable loss calculation, reduces interference from human factors, and improves calculation accuracy. It can be directly combined with thermal circuit models in engineering to evaluate cable core temperature in real time and dynamically determine current-carrying capacity. Attached Figure Description

[0019] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Figure 1 This is a flowchart of the intelligent calculation method for electromagnetic loss of multi-loop cable groups in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the model training and testing process in an embodiment of the present invention. Figure 3 This is a schematic diagram showing the connection of the intelligent computing device for electromagnetic loss of multi-loop cable groups in an embodiment of the present invention. Figure 4 This is a schematic diagram of the circuit laying of 8 single-circuit single-core cables (circuit 1 to circuit 8) in an embodiment of the present invention; Figures 5a to 5h The figures show the electromagnetic loss prediction error results for the eight single-loop single-core cables (loop 1 to loop 8) in the embodiments of the present invention. Detailed Implementation

[0020] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and, together with the embodiments of the present invention, serve to illustrate the principles of the present invention.

[0021] Example 1 One embodiment of the present invention discloses an intelligent calculation method for electromagnetic losses in multi-loop cable groups, such as... Figure 1 As shown, it includes: Step S1: Establish an electromagnetic field model of a cable group containing multiple loops using the finite element method, randomly generate current samples and calculate the corresponding cable core and metal sheath losses to form a current-loss sample set. Step S2: Construct the GWO-RBF-Adaboost intelligent prediction model. Use the Grey Wolf Optimization Algorithm (GWO) to globally optimize the center point, width, and output weights of the Radial Basis Network (RBF). Use the RBF neural network as the base learner of Adaboost and train the current-loss mapping relationship model using the current-loss sample set. Step S3: Input the current data of the multi-circuit cable group to be evaluated into the trained GWO-RBF-Adaboost model to predict the core loss and sheath loss of each circuit.

[0022] Specifically, in step S1, the electromagnetic field model of the cable group containing ≥3 loops is established. The Poisson equation is used for the conductive region and the Laplace equation is used for the non-conductive region. The boundary condition at infinity is applied. The eddy current loss of the corresponding cable core and metal sheath is calculated by randomly generating multiple sets of current input samples, forming a current-loss sample set.

[0023] Preferably, when building the electromagnetic field model of the cable group, the cable core and the metal sleeve are set as single coils, and each loop cable core is input with three-phase AC current with a phase difference of 120 degrees. According to the cable grounding method, when single-end grounding, the current of the metal sleeve is set to 0, and when double-end grounding, the metal sleeve is not set separately.

[0024] More specifically, step S1 includes: Step S101: Establish vector magnetic potential equations for each region of the cable field, with vector magnetic potential as the independent variable, including the Poisson equation for the conductive region, the Laplace equation for the non-conductive region, and the boundary condition at infinity. Specifically, the vector magnetic potential equations for each region of the cable field are as follows: ; In the formula, is the Laplace operator, representing the second derivative in space; The imaginary unit; Angular frequency; Electrical conductivity; is the magnetic permeability.

[0025] Step S102: Introduce the Coulomb specification to obtain the finite element matrix equation, calculate the vector magnetic potential of each node, and derive the eddy current density of the cable core region composed of the conductor, insulation layer and metal shielding layer. The formula for calculating the eddy current density in the cable core region is as follows: ; In the formula, In three-phase alternating current, the first The vector magnetic potential of the phase.

[0026] Step S103: Calculate the unit electromagnetic loss; the unit electromagnetic loss includes the loss of the cable core region calculated based on the eddy current density and the source current density, and the loss of the metal sheath region calculated based on the eddy current density. The losses in the cable core region are calculated from the eddy current density and the source current density: ; In the formula, , These are the source current density and eddy current density, respectively. The area of ​​the cable core region; Losses in the metal sleeve region calculated based on eddy current density: ; In the formula, This represents the area of ​​the metal casing region.

[0027] Step S104: Randomly generate multiple sets of loop current values, repeat step S103 to calculate the cable core loss and metal sheath loss corresponding to each set of current values ​​as current-loss samples, and form a current-loss sample set.

[0028] Specifically, the GWO-RBF-Adaboost training process in step S2 includes: Step S201: Read in the current-loss training set; Step S202: Initialize the gray wolf population; each wolf corresponds to a set of RBF center, width, and output weights; Step S203: Perform iterative optimization; In each iteration, an RBF weak learner is generated using the current wolf parameters and the Adaboost prediction error is calculated. The wolf pack position is updated by α, β, and δ wolves with the reciprocal of the error as the fitness. At the same time, Adaboost adjusts the weights of the samples and the weak learner until the iteration or error termination condition is met. Step S204: Output the GWO-RBF-Adaboost model corresponding to the optimal wolf parameters.

[0029] Specifically, the iterative optimization process in step S203 includes: Step S2031: Calculate model error; Use the RBF network parameters corresponding to the current gray wolf position to train the Adaboost ensemble model, and calculate the prediction error on the training set; Step S2032: Calculate fitness; Calculate the fitness value of each gray wolf individual based on the model error. Generally, the smaller the prediction error, the higher the fitness. Step S2033: Determine whether the convergence condition is met. If not, proceed to step S2034; if yes, proceed to step S2035. Step S2034: Update the gray wolf's position and return to step S2031 to continue model training; Based on the hierarchical hunting mechanism of the gray wolf optimization algorithm, the positions of ω wolf are updated by α, β, and δ wolves, and the RBF network parameters are adjusted accordingly. Step S2035: Evaluate the model performance at the current location of the gray wolf, and record the individual errors and the overall prediction error after integration; Step S2036 determines whether the number of iterations or the convergence condition has met the requirements; if yes, output the GWO-RBF-Adaboost model corresponding to the optimal wolf parameters; otherwise, proceed to step S2037. Step S2037: Adjust sample weights according to the error; The Adaboost mechanism adjusts the sample weights according to the prediction error, so that the subsequent weak learner pays more attention to the samples that are difficult to fit, and then returns to step S2031.

[0030] More specifically, the update formula for updating the position of wolf ω, guided by α, β, and δ wolves, is as follows: ; in, ; ; ; This is the current position of Gray Wolf ω. This is the update location for Grey Wolf ω; , , Let α, β, and δ represent the current positions of the wolves. , , , , , Let be the synergy coefficient; where, ; ; ; ; ; ; It is a random vector; It decreases linearly from 2 to 0.

[0031] The Grey Wolf Optimization Algorithm is used to determine the center point, the width of the radial basis function, and the output weight matrix of the RBF neural network, and to find the optimal solution.

[0032] RBF neural networks have the characteristic that only a few connection weights in a local domain of the input space affect the output, thus they have the advantages of fast learning speed, good convergence, simple structure, and no tendency to get stuck in local extrema.

[0033] Furthermore, the RBF network parameters include: hidden layer center vector, the dimension of which is equal to the number of input current loops; expansion constant, which can be a single scalar or an independent scalar for each center; output weight matrix, the dimension of which is the number of centers × the number of output loss categories; and cable core loss and metal sheath loss are output separately.

[0034] AdaBoost is a classic ensemble learning algorithm. Its core lies in dynamically adjusting sample weights to optimize sample distribution. This means increasing the weight of samples that were misclassified in previous learners, combining multiple weak learners into a strong learner, and through iterative training, making subsequent weak learners pay more attention to previously misclassified samples, thereby improving overall performance.

[0035] Furthermore, the Adaboost mechanism adjusts sample weights based on prediction error, the process of which includes: 1) Calculate the normalized weighted error of the current weak learner. ; ; in For the first The weights of each sample; ; 2) Calculate the ensemble weights of the weak learner. ; ; in, The learning rate; 3) Update sample weights ; ; And for all Renormalization, This allows for greater focus on samples that are difficult to fit in the next iteration.

[0036] The specific model training and testing process flow in this embodiment is as follows: Figure 2 As shown.

[0037] In step S3, the current data of the multi-circuit cable group to be evaluated is input into the trained GWO-RBF-Adaboost model to predict the core loss and sheath loss of each circuit.

[0038] Furthermore, the predicted core loss and sheath loss of each circuit are used as heat sources for the thermal circuit model to calculate the cable core temperature, and to further evaluate the dynamic current carrying capacity and provide early warning of operating status.

[0039] In summary, the intelligent calculation method and device for electromagnetic losses of multi-circuit cable groups in this embodiment realizes convenient and rapid calculation of electromagnetic losses of multi-circuit cable groups. The established GWO-RBF-Adaboost model, through the synergistic effect of GWO, RBF and Adaboost, significantly improves the prediction accuracy, generalization ability and robustness of the model. The solution greatly improves the automation and intelligence level of cable loss calculation and reduces the interference of human factors. It can be directly combined with the thermal circuit model in engineering to evaluate the cable core temperature in real time and dynamically determine the current carrying capacity.

[0040] Example 2 This embodiment discloses an intelligent calculation device for electromagnetic losses in multi-circuit cable groups, used to implement the intelligent calculation method for electromagnetic losses in multi-circuit cable groups as described in Embodiment 1. Figure 3 As shown, it includes: The sample set construction module is used to establish an electromagnetic field model of a cable group containing multiple loops using the finite element method, randomly generate current samples and calculate the corresponding cable core and metal sheath losses to form a current-loss sample set. The model training module is used to construct the GWO-RBF-Adaboost intelligent prediction model. The center point, width and output weight of the radial basis function neural network RBF are globally optimized using the Grey Wolf Optimization Algorithm (GWO). The RBF neural network is used as the base learner of Adaboost, and the current-loss mapping relationship model is obtained by training with the current-loss sample set. The loss calculation module is used to input the current data of the multi-circuit cable group to be evaluated into the trained GWO-RBF-Adaboost model to predict the core loss and sheath loss of each circuit.

[0041] The more specific technical details and beneficial effects in this embodiment are the same as those disclosed in Embodiment 1. Please refer to them for details, and they will not be repeated here.

[0042] Example 3 This embodiment provides a specific example of an intelligent calculation method for electromagnetic losses in multi-loop cable groups to verify the effectiveness of the method.

[0043] In this embodiment, the implementation process of the method includes: A finite element electromagnetic field model of the cable group was constructed, with the cable core and metal sheath set as single coils. Each loop cable core was input with three-phase AC currents that were 120 degrees out of phase. According to the cable grounding method, when the grounding was single-end, the current of the metal sheath was set to 0, and when the grounding was double-end, the metal sheath was not set separately.

[0044] 149 sets of input currents were randomly generated, and 149 sets of cable core loss and metal sheath loss sample data were calculated to form a current-loss sample set.

[0045] Eighty current-loss samples were selected to form a training set, and the GWO-RBF-Adaboost model was trained to form a defined loss calculation model based on current.

[0046] A test set of 69 current-loss samples was selected to verify the effectiveness and correctness of the GWO-RBF-Adaboost model. Repeatedly train and test the model to obtain the final model that calculates the loss based on the current.

[0047] Based on the obtained model, the loss calculation of multi-circuit cable groups in cable operation and maintenance is applied.

[0048] In a specific example model, a cable bridge contains 8 single-core cables, and there is a strong mutual influence between the cables. The schematic diagram of the 8-circuit single-core cable laying is shown below. Figure 4 As shown, the electromagnetic loss of an 8-circuit single-core cable is intelligently calculated here.

[0049] 149 sets of input currents were randomly generated, as shown in Table 1. The corresponding electromagnetic losses were calculated using the finite element method, as shown in Tables 2-7. It should be noted that the first 10 data points are given in each table for illustrative purposes.

[0050] Table 1 Input Current Samples (A)

[0051] Table 2. Sample 1 of cable core loss results (W / m)

[0052] Table 3. Sample 2 of cable core loss results (W / m)

[0053] Table 4. Sample results of cable core loss - 3 (W / m)

[0054] Table 5 Results of metal sheath samples - 1 (W / m)

[0055] Table 6. Sample 2 of metal sheath loss results (W / m)

[0056] Table 7 Sample 3 of metal sheath loss results (W / m)

[0057] Eighty sets of current and loss data were selected as training samples to obtain a current-loss calculation model, and the remaining 69 sets of current-loss data were used as test samples. The error of the electromagnetic loss result of the 8-loop single-core cable calculated from the 69 sets of test samples is as follows: Figures 5a to 5h As shown, Figures 5a to 5h Corresponding to loops 1 to 8 respectively, the maximum error in each diagram does not exceed 0.25W / m.

[0058] Through testing and verification, the established GWO-RBF-Adaboost model significantly improves the model's prediction accuracy, generalization ability, and robustness through the synergistic effect of GWO, RBF, and Adaboost.

[0059] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart calculation method for electromagnetic loss of multi-circuit cable groups, characterized in that, include: Step S1: Establish an electromagnetic field model of a cable group containing multiple loops using the finite element method, randomly generate current samples and calculate the corresponding cable core and metal sheath losses to form a current-loss sample set. Step S2: Construct the GWO-RBF-Adaboost intelligent prediction model. Use the Grey Wolf Optimization Algorithm (GWO) to globally optimize the center point, width, and output weights of the Radial Basis Network (RBF). Use the RBF neural network as the base learner of Adaboost and train the current-loss mapping relationship model using the current-loss sample set. Step S3: Input the current data of the multi-circuit cable group to be evaluated into the trained GWO-RBF-Adaboost model to predict the core loss and sheath loss of each circuit.

2. The intelligent calculation method for electromagnetic loss of multi-circuit cable groups according to claim 1, characterized in that, In step S1, the electromagnetic field model of the cable group containing ≥3 loops is established. The Poisson equation is used for the conductive region and the Laplace equation is used for the non-conductive region. The boundary condition at infinity is applied. The eddy current loss of the corresponding cable core and metal sheath is calculated by randomly generating multiple sets of current input samples, forming a current-loss sample set.

3. The intelligent calculation method for electromagnetic loss of multi-circuit cable groups according to claim 2, characterized in that, Step S1 includes: Step S101: Establish vector magnetic potential equations for each region of the cable field, with vector magnetic potential as the independent variable, including the Poisson equation for the conductive region, the Laplace equation for the non-conductive region, and the boundary condition at infinity. Step S102: Introduce the Coulomb specification to obtain the finite element matrix equation, calculate the vector magnetic potential of each node, and derive the eddy current density of the cable core region composed of the conductor, insulation layer and metal shielding layer. Step S103: Calculate the unit electromagnetic loss; the unit electromagnetic loss includes the loss of the cable core region calculated based on the eddy current density and the source current density, and the loss of the metal sheath region calculated based on the eddy current density. Step S104: Randomly generate multiple sets of loop current values, repeat step S103 to calculate the cable core loss and metal sheath loss corresponding to each set of current values ​​as current-loss samples, and form a current-loss sample set.

4. The intelligent calculation method for electromagnetic loss of multi-circuit cable groups according to claim 1, characterized in that, The GWO-RBF-Adaboost training process includes: Step S201: Read in the current-loss training set; Step S202: Initialize the gray wolf population; each wolf corresponds to a set of RBF center, width, and output weights; Step S203: Perform iterative optimization; In each iteration, an RBF weak learner is generated using the current wolf parameters and the Adaboost prediction error is calculated. The wolf pack position is updated by α, β, and δ wolves with the reciprocal of the error as the fitness. At the same time, Adaboost adjusts the weights of the samples and the weak learner until the iteration or error termination condition is met. Step S204: Output the GWO-RBF-Adaboost model corresponding to the optimal wolf parameters.

5. The intelligent calculation method for electromagnetic loss of multi-loop cable groups according to claim 4, characterized in that, The iterative optimization process includes: Step S2031: Calculate model error; Use the RBF network parameters corresponding to the current gray wolf position to train the Adaboost ensemble model, and calculate the prediction error on the training set; Step S2032: Calculate fitness; calculate the fitness value of each individual gray wolf based on the model error. The smaller the error, the higher the fitness. Step S2033: Determine whether the convergence condition is met; if not, proceed to step S2034; if yes, proceed to step S2035. Step S2034: Update the gray wolf's position and return to step S2031 to continue model training; Based on the hierarchical hunting mechanism of the gray wolf optimization algorithm, the positions of ω wolf are updated by α, β, and δ wolves, and the parameters of the RBF network are adjusted accordingly. Step S2035: Evaluate the model performance at the current location of the gray wolf, and record the individual errors and the overall prediction error after integration; Step S2036 determines whether the number of iterations or the convergence condition has met the requirements; if yes, output the GWO-RBF-Adaboost model corresponding to the optimal wolf parameters; otherwise, proceed to step S2037. Step S2037: Adjust sample weights according to the error; The Adaboost mechanism adjusts the sample weights according to the prediction error, so that the subsequent weak learner pays more attention to the samples that are difficult to fit, and then returns to step S2031.

6. The intelligent calculation method for electromagnetic loss of multi-loop cable groups according to claim 5, characterized in that, The update formula for updating the position of wolf ω, guided by wolves α, β, and δ, is: ； in, ； ； ； This is the current position of Gray Wolf ω. This is the update location for Grey Wolf ω; , , Let α, β, and δ represent the current positions of the wolves. , , , , , Let be the synergy coefficient; where, ； ； ； ； ； ； It is a random vector; It decreases linearly from 2 to 0.

7. The intelligent calculation method for electromagnetic loss of multi-circuit cable groups according to claim 6, characterized in that, The Adaboost mechanism adjusts sample weights based on prediction error. The process includes: 1) Calculate the normalized weighted error of the current weak learner; ； in For the first The weights of each sample; ； 2) Calculate the ensemble weights of the weak learner; Weak learner weights , The learning rate; 3) Update the sample weights; ； And for all Renormalization, This allows for greater focus on samples that are difficult to fit in the next iteration.

8. The intelligent calculation method for electromagnetic loss of multi-circuit cable groups according to any one of claims 1-7, characterized in that, When building the electromagnetic field model of the cable group, the cable core and the metal sleeve are set as single coils, and each loop cable core is input with three-phase AC current with a phase difference of 120 degrees. According to the cable grounding method, when the single end is grounded, the current of the metal sleeve is set to 0, and when the double end is grounded, the metal sleeve is not set separately.

9. The intelligent calculation method for electromagnetic loss of multi-circuit cable groups according to claim 8, characterized in that, The predicted core loss and sheath loss of each circuit in step S3 are used as the heat source for the thermal circuit model to calculate the cable core temperature, and further to evaluate the dynamic current carrying capacity and provide early warning of the operating status.

10. A smart calculation device for electromagnetic losses of multi-circuit cable groups, used to implement the smart calculation method for electromagnetic losses of multi-circuit cable groups as described in any one of claims 1-9, characterized in that, include: The sample set construction module is used to establish an electromagnetic field model of a cable group containing multiple loops using the finite element method, randomly generate current samples and calculate the corresponding cable core and metal sheath losses to form a current-loss sample set. The model training module is used to construct the GWO-RBF-Adaboost intelligent prediction model. The center point, width and output weight of the radial basis function neural network (RBF) are globally optimized using the Grey Wolf Optimization Algorithm (GWO). The RBF neural network is used as the base learner of Adaboost, and the current-loss mapping relationship model is obtained by training with the current-loss sample set. The loss calculation module is used to input the current data of the multi-circuit cable group to be evaluated into the trained GWO-RBF-Adaboost model to predict the core loss and sheath loss of each circuit.

Images